Research On Intelligent Cooperative Voltage Regulation And Economic Operation Strategy Of New Distribution Network

Due to the rapid development of renewable energy generation and power electronics technology,high proportion of renewable energy and high proportion of power electronics have become important features of modern new power systems.As an important part of the new power system,the new distribution network is the key infrastructure to achieve a high proportion of renewable energy consumption and achieve the goal of carbon peaking and carbon neutrality.However,with the widespread integration of high proportion of renewable energy generation and diversified loads,there are many challenges to the safe,high-quality,and economical operation of new distribution networks.The mismatch between the source and load terminals,as well as the bidirectional flow of power flow,have led to more prominent issues of voltage exceeding limits in the distribution network system;The randomness of renewable energy generation and the spatio-temporal coupling of source-load make it more difficult for multi-microgrids in distribution network to supply each other and operate economically.Therefore,this thesis focuses on the optimization of new distribution networks under the high proportion of renewable energy access,and the main research content is as follows:(1)The multi-resource cooperative voltage regulation problem of high proportion of renewable energy in distribution network is studied.Firstly,considering the energy storage system,flexible load,static reactive compensator and a large number of photovoltaic power generation,a multi-node active and reactive cooperative voltage regulation optimization problem is established.Secondly,considering that the distribution network system model is difficult to obtain accurately and there are uncertainties in large-scale new energy and load,solving this optimization problem is extremely challenging.Therefore,the established optimization problem is reconstructed as Markov game,each controllable source-load device is designed as a controllable agent,and the multi-agent extensible neural network architecture is constructed based on the attention mechanism.Finally,an evolutionary multi-agent deep reinforcement learning method combining population evolution,curriculum learning and multi-agent deep reinforcement learning is proposed to realize collaborative voltage regulation between large-scale photovoltaic nodes and multiple resources.The proposed method can train the voltage regulation strategy suitable for large-scale photovoltaic power generation node coordination without knowing the precise model of distribution network and the prior information of uncertain parameters.(2)The collaborative participation of multiple types of hybrid equipment in voltage regulation of distribution network is studied.Firstly,considering the challenges in the distribution system,such as the difficulty of efficient collaboration of discrete,continuous,and hybrid devices at different time scales,the multi-time scale voltage regulation and network loss minimization optimization problem of multi-type hybrid devices collaboration is established.Secondly,it is particularly difficult to solve the optimization problem directly due to the constraints of distribution system model information difficulty to obtain accurately,multi-type hybrid equipment coupling in time and space,and uncertainty of source and load fluctuations.Therefore,according to the different time scales of control equipment,the optimization problem is constructed as a fast and slow-timscale Markov game.Finally,a hierarchical multi-agent deep reinforcement learning algorithm is proposed to solve the Markov game of the upper and lower layers.Finally,a voltage control strategy based on the cooperation of multiple types of hybrid devices is obtained.The proposed method can realize the coordination of discrete devices on slow-timescale,continuous devices on fast-timescale,and hybrid devices on fast and slow-timescale,which can control the bus voltage within the safe range and reduce the network power loss of the system.(3)The cooperative economic operation of grid-connected multi-microgrid systems under multi-stakeholder conditions is studied.Firstly,considering the coordination of various resources,such as source-network-charge-storage-computing,a multi-microgrid system with fog-assisted services was constructed to minimize the operation cost.Secondly,it is difficult to solve the optimization problem directly because of the strong coupling between multiple resources,the information privacy between multi-microgrids,and the uncertainty of source load resources.Therefore,the above problems are transformed into Markov game and each microgrid is constructed as an agent.Finally,a multi-microgrid energy management method based on multi-agent deep deterministic policy gradient is proposed to realize the cooperative economic operation of multimicrogrid systems.The proposed method can support the real-time decision-making of each microgrid and effectively improve the operation economy of multi-microgrid system without knowing the prior knowledge of uncertain parameters of microgrid and other private information of microgrid.(4)The problem of hydrogen microgrid participating in cooperative voltage regulation and economic operation of distribution network is studied.Firstly,the optimization problem of hydrogen microgrid participating in the voltage regulation of distribution network and the operation cost minimization of hydrogen microgrid is established.Secondly,due to the stochastic uncertainty of source and load resources,it is difficult to accurately obtain the model information of distribution network system,and there are spatio-temporal coupling constraints among multiple resources,so it is difficult to directly solve the optimization problem.Therefore,the optimization problem is further modeled as Markov game with heterogeneous agents,and each microgrid and controllable device are constructed as different agents.A voltage regulation and economic operation method based on attention multi-agent deep reinforcement learning is proposed.Finally,in order to improve the interpretability of the neural network-based strategy model,an interpretable multi-agent deep reinforcement learning algorithm based on the experience iteration and imitation learning framework is proposed to realize the interpretability of the voltage regulation strategy.The proposed algorithm achieves voltage regulation of a new distribution network and economic operation of hydrogencontaining microgrids without the need to know accurate model information and uncertain parameter prior information of the system.